Skip to content

v0.0.15

Choose a tag to compare

@G-OTW G-OTW released this 04 Oct 17:01
· 1 commit to master since this release

[0.0.15] - 2026-10-04

Added

  • Editor and Backtest: opt-in versioning of files and saved strategies: save a dated version with a note, preview, restore or delete it; a deleted strategy can keep its history.
  • MCP: agents can save, list, read, restore and delete versions of files and strategies, and switch versioning per item.
  • Fundamentals: new module with macro series from 13 public sources, company statements, filings, insider trades and transcripts, ETFs, a market calendar and alternative data, stored locally.
  • Fundamentals: stored series and followed companies refresh on their own and notify new data; companies, series and filings show in search and are readable by agents.
  • Fundamentals: choose the order providers are tried in for each dataset, and pages skip a provider that just refused or ran out of quota until you press Refresh.
  • Dashboard: Fundamentals widgets for a macro series or board, a company snapshot, a statement line over time, your companies, latest filings, upcoming earnings and valuation against stored peers.
  • Dashboard: Quant volatility and regime widgets, dataset and compatible-basket selection, and configurable seasonality metrics and VaR confidence.
  • Agent: dump a conversation to the Editor in a folder you pick, with dates, persona and model, reasoning in a quote, and optional tool calls and tokens.
  • Agent: the Financial Analyst studies companies from Fundamentals (statements, ratios, filings, transcripts); reset the persona to pick it up.
  • Data broker: SEC EDGAR, FRED, FMP, Finnhub, Quiver Quant, USPTO, LDA.gov and the public statistics providers are connectors; Alpha Vantage, EODHD and Massive keys serve fundamentals too.
  • Backtest: limit orders for entries and signal exits, a lower-timeframe check that downloads only the candles it needs, a separate maker fee, and bid/ask fills fetched for the bars that fill.
  • Paper trading: stops, targets and limit orders trigger on the live price with an immediate alert, signals stay on closed candles, and a provider without a live price is named.
  • Historical Data: bid/ask from Capital.com, OANDA, Interactive Brokers, FOREX.com, Alpaca and Massive, with a table of what each provider offers in the docs.
  • Paper trading: each session has a full-page dashboard with a back button, live prices, its book, a chart of entries and exits on the strategy timeframe, and the backtest analytics including 3D scatter and streaks.
  • Backtest: holes in the data are flagged, trades crossing one are excluded from the results, and one click downloads the missing candles.
  • Backtest: a price tick that rounds every fill and level against the trader, negative fees for rebates, and trading-window time zones that follow daylight saving.
  • Fundamentals: Company, ETF and Alternative data share a symbol picker with favourites and recently opened symbols.

Changed

  • Editor and Backtest: versions open in a searchable window, the current one is marked, a restore is saved as a new version, and notes are capped at 500 characters.
  • MCP: agents can read and switch the editor and strategy versioning settings.
  • MCP: module listings no longer carry body schemas; an agent fetches one endpoint's schema when it needs it, so large modules like Backtest are listed in full.
  • Dashboard: Fundamentals and Quant widgets show units, reporting basis, history coverage and sources, refresh stored results every five minutes, and adapt tables and accessible heatmaps to narrow cards.
  • Backtest: grid and DCA strategies only show the filters, sizing and advanced settings they actually use.
  • Backtest: stops and targets are tested on the candle a trade opens in, the take profit fills as a limit, and a trade closed mid-candle no longer reopens at that candle's open.
  • Backtest: a neutral grid trades long below its centre line and short above it, entry fees leave the cash at the fill, and Kelly sizing resumes once the edge comes back.
  • Backtest: settings with no meaning and malformed candles are refused before the run, naming the problem.

Fixed

  • Dashboard: Quant correlation reads the nested API matrix; statement growth matches the same fiscal period a year earlier, and negative statement bars extend below zero.
  • Dashboard: earnings can use a stored calendar when a company snapshot is missing; statement labels keep their space, and phone layouts stack cards without horizontal page overflow.
  • Search: content titles no longer fail to load when Resources is not among the searched modules.
  • Backtest: ATR stops and targets use the ATR of the last closed candle, no longer that of the entry candle, whose range is not yet known when the trade opens.
  • Backtest: position size, pyramid adds, margin, exposure caps and circuit breakers use what is known at the candle's open, no longer its close.
  • Backtest: no new trade opens at the open of a candle whose close an expired exit limit closed the previous trade at.
  • Paper trading: an entry is alerted at the close that signals it, a market order then fills on the first live price and a limit order is confirmed when it fills.
  • Backtest: a lower-timeframe or bid/ask download stops after the provider refuses it, and the run's warnings say so.
  • Backtest: no decision fills at the close that made it: exit when the entry stops holding, and an expired exit limit, now leave at the next open.
  • Backtest: a pyramid add fills before the candle's stop, which then closes the larger position and is tested on the stop the add moved.
  • Backtest: portfolio slots, sizing and caps see only what is known at each asset's own open, sessions opening later included.
  • Backtest: grid buys only below the last close, fills gaps at the open, stops at its level, flattens at the window's open, and without bounds uses the range known so far.
  • Backtest: the optimizer and sweep rank on the in-sample part under a split, and the multiple-testing haircut uses the data's real candles per year.
  • Backtest: stocks run on split and dividend adjusted prices where the provider has them, with a raw prices option and a warning otherwise.
  • Backtest: take profit can require the price to trade through it, the trailing stop and MAE/MFE ignore prices from before a fill or after an exit, and tick bid/ask uses the opening spread.
  • Paper trading: announced orders are sized on marked equity and carry their amount, for brokers that take notional orders, shown in the dashboard's orders tab.
  • Backtest: a grid candle that opens past one circuit stop exits at that stop, even when it also reaches the other.
  • Backtest: DCA MAE and MFE are positive amounts that count the contract multiplier once.
  • Backtest: the warm-up counts indicators left at their default length and MACD's signal line.
  • Backtest: an EMA or WMA chained on a series with gaps leaves the gaps empty instead of averaging zeros.
  • Backtest: Kelly sizing no longer counts a breakeven trade as a win.
  • Backtest: reverse side mirrors oscillator levels (RSI below 30 becomes above 70) and keeps ADX, ATR and volume filters unchanged.
  • Historical Data: the candle still forming is never stored, and a last candle saved before its period ended is read again on the next update.
  • Backtest and paper trading: no position is carried across missing data, and entries wait for indicators to warm up again after a hole.
  • Paper trading: a session downloads the candles missing after a pause by itself, and holes the provider has no data for are reported, not retried.
  • Backtest: the Markdown export no longer multiplies circuit-breaker thresholds by 100.